Verified data · as of Oct 28, 2026
Do AI trading bots make money?
Here's every AI trading result we could verify from public chain and venue data, in one place: AI agents, AI-run vaults, AI trading apps, AI models trading real money, and what products advertise. No screenshots, no self-reported PnL.
Short answer: some do, most can't show it, and the ones you can check are a mixed bag. Most AI trading bots don't publish a trading record, so their returns can't be checked at all. Where the trades are public, here is what we measured:
- AI agents on prediction markets made a small trading profit: +9.1% across 321 active agents registered on-chain over 90 days, on $57,570 of bets, and before what they pay for AI predictions.
- Hyperliquid vaults that say they're run by AI: 2 of 6 beat BTC over the same dates and 1 fell more than 30% from a peak.
- People using AI trading apps on Hyperliquid: in 6 of 8 apps, users lost money overall over 30 days, and in most apps fewer than half the users were in profit.
- AI models trading real money in Nof1's Alpha Arena: 2 of 6 finished Season 1 with a profit; in Season 1.5, 7 of 8 lost.
- Advertised returns: 17 of the 18 performance claims we checked have no record anyone else can verify, and 6 are backtests or hypothetical.
Past results don't predict future results, and none of this is a recommendation.
AI agents on prediction markets: small gains, small stakes
Olas Predict trader agents are autonomous AI agents registered on-chain (3,875 so far). Each bets on prediction markets from its own wallet, so every bet and payout is public. Over the last 90 days:
- Omen (Gnosis): 277 active agents bet $55,481 on settled markets for a trading ROI of +9.3%. Of the 138 with enough activity to rank, 125 made money; the median ROI was +8.7%.
- Polymarket: 44 active agents bet $2,089 on settled markets for a trading ROI of +4.9%. Of the 19 with enough activity to rank, 16 made money; the median ROI was +7.3%.
Two things keep this from being a money machine: stakes are a few dollars a bet in thin markets, so it doesn't scale, and the profit leaves out what each agent pays for AI predictions. Agents that stopped trading drop out of the ranking, which flatters the share that made money.
Full ranking of AI agents on prediction markets · Leaderboard: AI agents
Hyperliquid vaults that say they're run by AI: a split
Every open Hyperliquid vault with $100k+ and 30+ days live whose name or leader description claims AI. Returns are ours, computed from Hyperliquid's public data and before the leader's 10% profit share. As of Oct 28, 2026.
| Vault (AI label is the leader's claim) | Return | BTC, same dates | Max drawdown |
|---|---|---|---|
| AceVault Hyper01 | +17.1% | −26.5% | −24.2% |
| Akka Hyper AI | +9.8% | −20.3% | −5.0% |
| FuturAI Labs - High Sharpe, Low Volatility | +7.9% | +11.5% | −0.6% |
| Stratwise Multi-Asset Public | +6.9% | +22.5% | −2.5% |
| AIQuantPulse | +0.3% | +13.6% | −10.8% |
| DailyTradeAI Low risk fell 30%+ from a peak | −67.3% | −22.1% | −90.3% |
Leaderboard: AI-labelled vaults · All Hyperliquid vault ratings
People using AI trading apps: most lost a little
Apps that run AI agents or an AI copilot on Hyperliquid tag every order with a builder code, and Hyperliquid publishes every fill. From Sep 28, 2026 to Oct 27, 2026, 1,561 users traded $24.2M through 8 such apps. Realized PnL is before exchange fees, and the apps mix autopilot, copilot and manual orders.
| App | Users | Users' realized PnL | Users in profit |
|---|---|---|---|
| TrueNorth | 283 | −$1,361 | 51% |
| Senpi | 831 | −$4,714 | 40% |
| DXAP | 233 | $24k | 34% |
| Wayfinder | 43 | −$1,080 | 42% |
| Midas | 80 | −$4,067 | 59% |
| Katoshi | 16 | $632 | 38% |
| Minara AI | 73 | −$237 | 47% |
| Starchild | 2 | −$1 | 0% |
Every AI trading app on Hyperliquid, compared · Leaderboard: AI apps
AI models trading real money: most lost
Nof1's Alpha Arena gave leading AI models real money to trade. Season 1: crypto perps on Hyperliquid, $10,000 per model, Oct 18 – Nov 3, 2025:
- Qwen3 Max: +22.3%
- DeepSeek Chat V3.1: +4.9%
- Claude Sonnet 4.5: between −42.0% and −30.8% (sources differ)
- Grok 4: between −57.9% and −45.3% (sources differ)
- Gemini 2.5 Pro: between −56.7% and −45.5% (sources differ)
- GPT-5: between −62.7% and −58.7% (sources differ)
In Season 1.5 (US stocks) Grok 4.20 (entered as the Mystery Model) won with +12.1% overall, and seven of the eight models finished with a loss. On forecasting benchmarks the best models still trail expert human forecasters.
Advertised returns: mostly unverifiable
We recorded 18 public performance claims from AI trading products: win rates, returns, "beats the market". 6 are backtests, simulations or hypothetical by the product's own description; 11 are the product's own numbers with no record a third party can check; 1 can be verified. A claim without evidence may still be true, but you can't tell.
How to check a bot's returns yourself
A public wallet, vault or exchange-verified record you can open yourself. Screenshots and dashboards the seller controls prove nothing.
A bot that made 20% while BTC made 40% lost you money compared with doing nothing. Always check the same dates.
Max drawdown tells you what you'd have had to sit through. A 90% drop needs a 900% gain to get back.
Subscription, exchange fees, the AI's own costs, and bots that quietly stopped: survivors make any list look better.
FAQ
Can an AI trading bot make you money?
Some make a little and many lose. Of the AI trading results anyone can verify, registered AI agents on prediction markets made a small trading profit on small stakes, AI-labelled Hyperliquid vaults split, and users of most AI trading apps on Hyperliquid lost money overall. See the verified leaderboard for current numbers.
Are AI trading bot returns real?
Most can't be checked. Of 18 advertised claims we recorded, 17 have no independent record and 6 are backtests or hypothetical. Ask for a public account you can open yourself.
Which AI trading bot is the most profitable?
Nobody can answer that for bots without a public trading record, which is most of them. For the ones with public accounts, the verified leaderboard ranks them on risk-adjusted return after a drawdown gate, and the order changes as markets move.
Why aren't 3Commas, Cryptohopper or Telegram bots on the leaderboard?
Their results live inside each user's own exchange account, so there's no public record to verify. We track how much they're used on trading bot usage stats and what they can do with your money on AI trading bots.